The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands for repository management and file operations. Access is restricted via frontmatter configuration to specific commands: git, cd, cp, mkdir, and basename. This scoping prevents the execution of arbitrary or unauthorized system commands.
[DATA_EXFILTRATION]: The skill transmits implementation plans and task status metadata to an external management service using dedicated MCP tools (mcp__handoff__handoff_push_plan and mcp__handoff__handoff_sync_status). This data flow is central to the skill's purpose of synchronizing planning state between the local environment and a handoff coordinator.
[PROMPT_INJECTION]: The skill incorporates content from local project documentation (such as RESEARCH.md and DESCRIPTION.md) into its planning context. While this introduces an indirect prompt injection surface common to agentic tools, the skill uses structured templates and explicit task creation tools to maintain control over the generated output.

aif-plan